Welcome

Last updated 3 months ago

Welcome

AXAR AI is a lightweight framework for building production-ready agentic applications using TypeScript. It’s designed to help you create robust, production-grade LLM-powered systems using familiar coding practices—no unnecessary abstractions, no steep learning curve.

⌾ Yet another framework?

Most agent frameworks are overcomplicated. They prioritize flashy demos over practical use, making it harder to debug, iterate, and trust in production. Developers need tools that are simple to work with, reliable, and easy to integrate into existing workflows.

At its core, AXAR is built around code. Writing explicit, structured instructions is the best way to achieve clarity, control, and precision—qualities that are essential when working in the unpredictable world LLMs.

If you’re building real-world AI applications, AXAR gets out of your way and lets you focus on shipping reliable software.

🌍 Hello world!

Here's a simple example of an AXAR agent:

import { model, systemPrompt, Agent } from '@axarai/axar';

// Define the agent.
@model('openai:gpt-4o-mini')
@systemPrompt('Be concise, reply with one sentence')
export class SimpleAgent extends Agent<string, string> {}

// Run the agent.
async function main() {
  const response = await new SimpleAgent().run(
    'Where does "hello world" come from?',
  );
  console.log(response);
}

main().catch(console.error);

Why use AXAR

🧩 Type-first design: Structured, typed inputs and outputs with TypeScript (Zod or @annotations) for predictable and reliable agent workflows.
🛠️ Familiar and intuitive: Built on patterns like dependency injection and decorators, so you can use what you already know.
🎛️ Explicit control: Define agent behavior, guardrails, and validations directly in code for clarity and maintainability.
🔍 Transparent: Includes tools for real-time logging and monitoring, giving you full control and insight into agent operations.
🪶 Minimalistic: Lightweight minimal design with little to no overhead for your codebase.
🌐 Model agnostic: Works with OpenAI, Anthropic, Gemini, and more, with easy extensibility for additional models.
🚿 Streamed outputs: Streams LLM responses with built-in validation for fast and accurate results.

Inspirations

Jump right in

NextQuickstart